laliteshpawar.com/
Back to What I Learn
technical breakdown · last crawled: this week · source: prompt & output audits

LLM retrieval & citation patterns

Index Progress & Audit Level
74%
Mapping how vector embeddings and semantic proximity influence whether AI tools cite enterprise documentation.

This technical breakdown documents key findings from running prompt audits and structured entity schema tests across enterprise search workloads.

Core Technical Hypotheses

Vector distance threshold decreases by 18% with structured JSON-LD.
LLMs favor direct Q&A headers over dense marketing narrative.

Implementation Snippet

schema-example.jsonld
{
  "@context": "https://schema.org",
  "@type": "TechArticle",
  "name": "LLM retrieval & citation patterns",
  "author": {
    "@type": "Person",
    "name": "Lalitesh Pawar"
  }
}